Quantification of the TMS-EEG response in epilepsy

(1)

Maria Inês Fonseca Silva Santos

Licenciatura em Ciências de Engenharia Biomédica

Quantification of the TMS-EEG response in

epilepsy

Dissertação para obtenção do Grau de Mestre em Engenharia Biomédica

Orientadores :

Esther ter Braack, M.Sc., Universiteit Twente,

Países Baixos

Michel J.A.M. van Putten, Prof. Associado,

Universiteit Twente, Países Baixos

Co-orientadora :

Carla Quintão Pereira, Profª. Auxiliar, FCT - UNL

Júri:

Presidente: Profª. Doutora Maria Adelaide de Almeida Pedro de Jesus Arguente: Prof. Doutor Mário António Basto Forjaz Secca

(2)

(3)

Quantification of the TMS-EEG response in epilepsy

(4)

(5)

(6)

(7)

Acknowledgements

As I am writing the last few pages of this thesis, I come to thecliché, but nonetheless valid, realization that five years have gone by in a blink. I have been privileged to complete my higher education and now can carry on into the world with knowledge that will surely assist me in in my adversities and victories.

I am truly grateful to my university, Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa for having been my home for the past five years. The spirit given by the campus and the environment is ideal for fostering young minds that will develop thoughts to change the world.

For the development of this thesis, I had the wonderful supervision of Esther ter Braack and Michel van Putten from the University of Twente. Esther, your daily assistance and guidance during the development of this thesis was truly essential, thank you. Michel, I appreciate you welcoming me into your research group. This was a great experience. I thank Carla Quintão for remaining my link to FCT-UNL and for all the help provided in the seven months abroad.

Also at the University of Twente, I express my gratitude to many of the Ph.D. and M.Sc. students who worked by my side throughout the past seven months. This experience was unforgettable, and part of it was also due to the friends from around the world that gathered for the Spring semester at UT in 2012. To my Portuguese family away from home, thank you and I miss you.

For all the teachers I’ve had over the past five years and even before, I show true appreciation because it is your passion for teaching that has allowed me to gather information about so many diverse topics and use those thoughts to transform my own ideas. However, this time would have seemed longer if not for the companionship of classmates and friends that have stuck by me through assignments, exams, parties and the overall university experience. Angela Pimentel, Rodolfo Abreu, Ricardo Alves, Ana Cartaxo, Diliana Rebelo, Ana Rita Pereira, Cátia Q., Margarida Félix, Ana Teresa Neves, Sofia Dias, Mafalda Correia, and Inês Santos, you have made this part of my life more exciting and memorable. To my fellow BESTies I say thank you for all the experiences and memories.

I cannot forget to mention those who have been by my side for so long though it still feels like yesterday that we met. You know who you are, Diana Farcas, Irene Esteban, João Quintas, Catarina Stewart, Sarah Roberts, and Audi Rogers. Thank you for just being there and for theget-togethersevery once in a while.

(8)

we are today. That is why I acknowledge those friends that I have met throughout my life and that are now spread out all over the world.

(9)

“Efforts and courage are not enough without purpose and direction.”

(10)

(11)

Abstract

Purpose: The purpose of this thesis was to provide quantitative measures of the co-registration of transcranial magnetic stimulation (TMS) and electroencephalogram (EEG). The EEG is used to study changes in the neuronal activity evoked by the non-invasive technique TMS. These effects are determined mainly based on clinical judgment. Current uses in the diagnosis of epilepsy are based only on EEG, not taking into consideration the low sensitivity in the interictal period, in particular if routine recordings are used.

Methods: Patient data was gathered, analyzed and compared to healthy controls. A total of ten patients and eighteen healthy subjects underwent sessions of 75 TMS pulses. The responses to the pulses were filtered and averaged. The use of topographical scalp plots of amplitude and power, and time-series analysis of power in search for late responses provide results which enable separation of epilepsy patients and healthy controls. By investigating the significance of the results it is also possible to determine, in a quantitative way how reliable the methods are for distinguishing between the two groups.

Results: The definition of what is a response is critical in this project, and as such must consider: significant power change, be above a certain amplitude, and be localized. Still, this procedure results in a non distinguishable threshold to separate both groups.

Conclusions: Analysis of the receiver operating characteristic (ROC) curves also led to the understanding the method established is not entirely reliable because it cannot in fact determine differences. Since all patients were under treatment with anti-epileptic drugs (AEDs), it becomes necessary to elaborate a pilot study with recently diagnosed subjects where hyperexcitability is still present.

(12)

(13)

Resumo

Objectivo: O objectivo deste trabalho contempla fornecer medidas quantitativas do registo simultâneo da estimulação magnética transcraniana (TMS, do inglêstranscranial magnetic stimulation) e do electroencefalograma (EEG). O EEG é utilizado para estudar alterações na actividade neuronal evocada pela técnica não-invasiva TMS. Estes efeitos são determinados predominantemente com base na avaliação clínica. A utlização corrente no diagnóstico da epilepsia é baseada apenas no EEG, não tendo em consideração a sua baixa sensibilidade no período interictal, especialmente em procedimentos de rotina.

Métodos: A informação de pacientes foi recolhida, analisada e comparada a controlos saudáveis. O total de dez pacientes e dezoito pessoas saudáveis foram sujeitos a sessões de 75 pulsos a cujas respostas foram aplicados filtros e obtida a sua média. A utilização da representação topográfica do escalpe em amplitude ou potência, e a análise em tempo da potência, na procura de respostas tardias, providenciam resultados que permitem a separação entre pacientes epilépticos e controlos saudáveis. Ao investigar a significância dos resultados é também possível determinar, de uma forma quantitativa, o quanto os métodos são fiáveis para distinguir entre os dois grupos.

Resultados: Em que consiste uma resposta é uma definição crítica para este projecto, e para tal é necessário considerar: alterações de potência significativas, ser acima de uma certa amplitude, e ser localizada. Este procedimento leva a um nível de separação pouco distinto entre os dois grupos.

Conclusões:A análise das curvas ROC (do inglêsreceiver operating characteristic) também conduz a uma compreensão de que o método estabelecido não é inteiramente fiável uma vez que não consegue determinar diferenças. Visto que todos os pacientes estão sujeitos a tratamento com medicamentos anti-epilépticos (AEDs do inglês anti-epileptic drugs), torna-se necessário elaborar um estudo piloto com indivíduos recentemente diagnosticados e onde a hiperexcitibilidade ainda se encontra presente.

(14)

(15)

List of Figures

2.1 Number of people with epilepsy in WHO regions, where N is the number of responding countries in each area. The numbers (N=105) are only based on information provided by respondents to WHO’s Atlas. These were not corrected for those countries that did not respond. . . 6 2.2 Most frequently reported causes of epilepsy, as reported by countries part

of WHO (N=149) . . . 7 2.3 Characteristic EEG rhythms - delta, theta, alpha, and beta. As defined in

table 2.1 . . . 9 2.4 Galvani’s experiment that pioneered the subject of electrophysiology. This

involved the study of muscular contraction in a frog by touching its nerves with electrostatically charged metal . . . 10 2.5 Diagram of recording a single EEG channel. The differential amplifier

measures the potential between two electrodes, where one of them is treated as the reference. The second amplifier prepares the signal for AD conversion and storage (lower path). Before the development to digital, EEG was stored on folded paper (upper path) . . . 11 2.6 24 seconds of EEG around an epilepsy seizure onset for a patient of a study

conducted by Lantzet al.. The vertical bar indicates a visually estimated seizure onset . . . 12 2.7 An example of generalized slow spike-wave complexes (around 2 s) found

in a child with severe epileptic seizure disorder . . . 13 2.8 A circuit with a sinusoidal current pulse passing through the coil. A circuit

with a sinusoidal current pulse passing through the coil (L). A gate signal from the stimulator opens the thyristor switch (S), thus discharging the high-voltage capacitor (C) through the coil . . . 16 2.9 Principles and chain of events in TMS. The current pulses in the coil

generates a magnetic fieldBthat, in turn, induces an electric fieldEthat is strongest near the coil . . . 17 2.10 TEP: TMS-evoked potential. There is a clear identification of the major

(18)

3.1 A Styrofoam head is used to simulate a subject, with the reflecting balls placed on a headband. In the background is a 3D model of a head constructed using MRI scans. . . 27 3.2 The shape of the TMS pulse used when acquiring the EEG data for every

individual . . . 28 3.3 The magnetic stimulator with the touch screen interface. The

robot-navigated system is used for accurate positioning of the figure-of-eight coil. . . 28 3.4 Representation of the 64 channels used to acquire the EEG signal.

Highlighted in green are the first 32 electrodes . . . 30 3.5 Auto-scaled image of the EEG signal, including the TMS pulses. In this

image, it will be necessary to select a threshold for detecting all the pulses. 31 3.6 This image represents the same EEG signal (blue) in the previous figure,

with an adapted scale and with the trigger (red) method identifying the peaks of TMS pulse. . . 32 3.7 TMS-evoked potential obtained for patient number 4 with stimulation site

MCR. The peaks P30, N45, P55, N100, and P180 are identified. . . 33 3.8 Example of topographic illustrations given by the developers of EEGLAB.

The several images indicate the time at which the calculation was made . 34 3.9 Identification of what was considered the baseline in the time-series, from

800 to 200 ms before the pulse. . . 35 3.10 Setting used to determine response changes (increase or decrease) in

patients and healthy controls. Electrodes FC2, FC4, FC6, C2, C4, C6, CP2, CP4, and CP6 highlighted in blue, while the corresponding opposite electrodes on the left side are in red. . . 35 3.11 Identification of the time intervals determined as the baseline period (800

to 200 ms before the pulse - orange), early response period (0 to 400 ms after the pulse - purple) and late response period (400 to 950 ms after the pulse - blue). . . 37 3.12 Use of the software of NeuroCenter Viewer to identify the trials which

contain artifacts and which will be removed from the follow-up analysis for healthy subject 17, stimulation site MCL. . . 38 3.13 Use of the software of NeuroCenter Viewer to identify the trials which

contain artifacts and which will be removed from the follow-up analysis for patient 7, stimulation site MCR. . . 39

4.1 EEG time-scale results obtained from 64 channels from an epilepsy patient, after baseline removal, interpolation, and filtering. Results from patient number 4 with stimulation site MCL. . . 42 4.2 Topographical plot showing amplitude with 4 or 5 ms intervals, from -10

to 68 ms. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Results from patient number 4 with stimulation site MCL. . . 43 4.3 Topographical plot showing amplitude with 50 ms intervals, from -10 to

(19)

4.4 Time-series result for channel Cz, from -200 to 400 ms with images from the topographical plots for certain time periods. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Results from patient number 1 with stimulation site MCR. . . 45 4.5 Power graphs which quantify the GMFA results obtained for all 61

channels in 8 patients. The baseline from 800 to 200 ms before the pulse is in red and the GMFA is in blue. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Results with stimulation site MCR. . . 46 4.6 Power graphs which quantify the GMFA results obtained for only 9

channels surrounding the stimulation site in 8 patients. The baseline from 800 to 200 ms before the pulse is in red and the GMFA is in blue. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Results with stimulation site MCR. . . 47 4.7 Number of channels in each interval of normalized power. Time interval is

from 0 to 1 second after the TMS pulse. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Sample of 8 patients with stimulation site MCR. . . 48 4.8 Number of channels in each interval of normalized power. Time interval

is from zero to one second after the TMS pulse. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Sample of 9 healthy subjects with stimulation site MCR. . . 49 4.9 Topographical plot showing the absolute power difference (post-stimulus

result subtracted by baseline) calculated using RMS. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. . . 50 4.10 Average signal consisting of epochs of two seconds, at channel C4

for patient number 2 with stimulation site MCR. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz. . . 52 4.11 Average signal consisting of epochs of four seconds, at channel C4

for patient number 2 with stimulation site MCR. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz. . . 52 4.12 Topographical plot showing the absolute power difference (post-stimulus

result subtracted by baseline) calculated using RMS. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 4 s. . . 53 4.13 Topographical plot showing the absolute power difference (post-stimulus

result subtracted by baseline) calculated using RMS. Several examples of both stimulation sites. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 4 s. . . 55 4.14 ROC curves constructed with the sensitivity and specificity results

obtained from the several thresholds for determining a response and thus separate patients from healthy controls. The blue line represents a p-value of 0.05 and the red line represents a p-value of 0.01. When only one line is visible, the results are the same for both p-values. . . 56

(20)

(21)

List of Tables

2.1 Common definition of frequency bands in the EEG. . . 9 2.2 Typical interictal epileptiform discharges found on the EEGs of patients

with characteristic epilepsy syndromes or etiologies . . . 12 2.3 Sharp transients representing normal EEG variants, being easily confused

with epileptiform discharges . . . 14 2.4 TMS parameters that can be obtained by its use in several forms depending

on the objective of the experiment . . . 18

3.1 Table built to establish the sensitivity and specificity of the results determined. D+ is the presence of the disease, D- is the absence. Test + indicates a positive test outcome, while Test - represents a negative test outcome. . . 39

4.1 Overview of the presence of late responses in epilepsy patients and healthy controls, using GMFA to calculate the power in 61 channels. Late responses are considered to exist around 800 ms. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. Results from stimulation site MCR. . . 48 4.2 Overview of the presence of late responses in epilepsy patients and healthy

controls. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 2 s. For p <0.01 and RMS values >|2| µV. . . 51 4.3 Overview of the presence of late responses in epilepsy patients and healthy

controls. Filter definitions are band-pass 1 to 80 Hz and band-stop 49 to 51 Hz, with epochs of 4 s. For p <0.01 and RMS values >|1| µV. . . 54 4.4 Overview of the presence of late responses in epilepsy patients and healthy

(22)

(23)

List of Abbreviations

AD Analog-to-digital

ADM Abductor digiti minimi

AED Anti-epileptic drug

C Central

EEG Electroencephalography

EMG Electromyography

EOG Electro-oculogram

EP Evoked potential

ERP Event related potential

F Frontal

Fp Frontopolar

FT Fourier transform

GABA Gamma-aminobutyric acid

GMFA Global mean-field amplitude

HFO High-frequency oscillation

IED Interictal epileptiform discharge

MCL Motor cortex left

MCR Motor cortex right

MEP Motor evoked potential

MRI Magnetic Resonance Imaging

MT Motor threshold

(24)

O Occipital

P Parietal

PCA Principal component analysis

PET Positron Emission Tomography

RMS Root mean square

ROC Receiver operating characteristic

SOZ Seizure onset zone

SPECT Single-Photon Emission Computed Tomography

SREDA Subclinical rhythmic electrographic discharges of adults

TEP TMS evoked potential

TMS Transcranial Magnetic Stimulation

TMS-EEG Combined use of TMS and multi-channel EEG

(25)

1

Introduction

1.1 Motivation

Since ancient times, epilepsy has been associated with evil and religious entities. For centuries it has been surrounded by fear and discrimination. Even though there is still some social stigma in certain regions, today, epilepsy is viewed as a neurological disturbance where a high number of nervous cells are excited simultaneously during a seizure. Epilepsy consists of more than seizures for the affected individual, especially because it leads to many interacting psychological, medical, economic, and social repercussions.

Given the current digital advances, it is surprising that major breakthroughs in the clinical use of quantitative electroencephalographic (EEG) analysis are somewhat limited. This situation contrasts with advances in, for example, neuroradiology, where digital signal analysis has greatly influenced imaging techniques. Along with the long learning curves associated with the visual interpretation of the EEG in a clinical environment, there can be several inter- and intra-observer inconsistencies. Furthermore, qualitative information may not always be suited to communicate particular features. Elements of spatio-temporal dynamics are often difficult to translate into the language domain. For this reason, it is necessary to develop alternative presentations which may assist the interpretation.

(26)

1. INTRODUCTION 1.2. Objective

the activation mechanisms of TMS. The second objective is to confirm the potential applications of the combined use of TMS and multi-channel EEG (TMS-EEG) as a tool for neurophysiological research and diagnostic purposes.

Functional brain mapping methods such as EEG, functional magnetic resonance imaging (MRI) and positron emission tomography (PET) have, so far, permitted the non-invasive investigation of the functional organization of the human brain by providing maps of the distribution of activity [1, 11]. However, functional MRI and PET have also made it difficult to investigate the dynamical connectivity between neurons in the brain and are of little use in determining cortico-cortical connections [6, 11] because of their low temporal resolution. For these reasons, they would not be adequate for the quantitative study of epilepsy. Using TMS with the mentioned neuroimaging methods expands the applicability of TMS to the study of cortical reactivity and connectivity. The temporal resolution shown by these methods does not permit the establishment of the time course activation of the stimulated area and remote sites [5]. EEG, however, has a very good temporal resolution (in the order of milliseconds), that combined with TMS could provide new information concerning diagnosis and therapies. TMS is different from otherin vivomethods that show the function of the human brain because instead of observing the brain in operation, neurons are actually triggered into action [1].

Responses to TMS-EEG can be defined as early and late responses. Early responses usually include most, if not all, of the TEP. An epoch surrounding a TMS pulse is defined from one second before to one second after the pulse. A study by Valentín et al. [12] identified late TMS-EEG responses in 73% of epilepsy patients (11 out of 15), whilst in 100% of healthy subjects there was no such response. The late response period defined in this study is from 100 to 1000 ms. The finding suggests that late responses are abnormal responses of the epileptic cortex to the TMS. This might indicate the existence of a hyper-excitable cortex under the stimulated area. The sample size was small but these preliminary results introduce the possibility of more certain and earlier diagnosis using the combination of EEG and TMS.

1.2 Objective

The objective of this project is to provide quantitative measures of the co-registration of TMS and EEG activity mapping that currently exists. To achieve this aim, signals will be processed and the tools which analyze and extract information from these signals will be developed. The attention is thus drawn to the EEG signals obtained after TMS. Through the use of topographical plots to evaluate potentials and determination of power spectrum, the amount of information from the collected data is reduced. The techniques developed and applied were used to create a clearer understanding and provide significant information regarding differences between epilepsy patients and healthy subjects. Individually or combined, they were applied to the signals collected. An analysis to evaluate the quality of these results was also done.

(27)

1. INTRODUCTION 1.3. Thesis Overview

1.3 Thesis Overview

The study evaluates the combined use of TMS-EEG in the development of quantitative tools for the diagnosis of epilepsy. From the data processing and analysis of eighteen healthy subjects and ten patients with epilepsy, there was an assessment of the occurrence and significance of the TMS-evoked responses. The implemented algorithms were developed using MATLAB, a high-performance interactive software ideal for scientific and engineering computation.

(28)

(29)

2

Theoretical framework

This chapter provides a framework for the main topic of this thesis. Some insights are given regarding the condition of epilepsy and some of its associated mechanisms. In regards to the techniques of EEG, TMS, and the combination of TMS and EEG, more detail is given on how they have evolved and how they are applied today. More focus is given to the co-registration of TMS and EEG due to its importance in the application of the quantification tools that were developed during this project.

2.1 Epilepsy

Rhythmic activity is a fundamental property of neural elements. Its organization is in the form of complex patterns which depend on the state of the brain and on the task that is being executed. Synchronization of oscillations across neuronal elements, either locally or over longer distances, is one of the organizing principles of rhythmic activity [14]. Prevailing rhythmicity and organization could be a sign of abnormality, and disorganized oscillations do not necessarily imply abnormality [15]. Brain oscillations have a range from 0.05 to 600 Hz, where fast wave activity is associated with the awake state and slower oscillations with sleep [16]. This demonstrates that oscillatory activity in distinct frequency bands has been related to specific functions.

(30)

2. THEORETICAL FRAMEWORK 2.1. Epilepsy

peak at the age of 65 and older [17].

Figure 2.1: Number of people with epilepsy in WHO regions, where N is the number of responding countries in each area. The numbers (N=105) are only based on information provided by respondents to WHO’s Atlas. These were not corrected for those countries that did not respond [18].

The pathophysiological basis for human epilepsy is thought to be a cortical imbalance between inhibitory and excitatory mechanisms involving increased, hyper-synchronous and autonomous activity [12, 19]. In other words, it is a short-lasting occurrence of signs due to the abnormal synchronization of neuronal activity in the brain. Brain cells produce electrical discharges through the use of chemical interactions. Each cell either excites or inhibits other brain cells with its discharges. When the balance shifts too much in the direction of excitation, then a possible outcome is a seizure. Hughlings Jackson, more than half a century before the discovery of the human EEG, defined epilepsy as an “occasional sudden, excessive rapid and local discharges of grey matter” [19].

In fact, epilepsy is known as a seizure disorder, classifying the seizure as the event and epilepsy as the disorder. The diagnosis is usually made after a person has had at least two spontaneous seizures [20] and if the brain has an increased tendency to generate seizures [21]. These seizures may be related to brain injuries or family tendency, also known as symptomatic epilepsy; but often (six out of ten) the cause is unknown, i.e. idiopathic epilepsy.

2.1.1 Types of Seizures

Seizures are, by convention, divided into two basic types: generalized and partial/focal. This division is based on both clinical and electrophysiological terms. The type of seizure depends on several factors, such as where the abnormal electrical discharge occurs in the brain. The temporal lobe is an area prone to generate seizures. Parts of the brain most commonly involved in adult epilepsy, such as the amygdala and the hippocampus, are found in the temporal lobe. Physiological events can range from disruption of the elementary functions of the brain region involved in a well localized discharge; alteration of consciousness and behavior in more widespread discharges [19].

(31)

2. THEORETICAL FRAMEWORK 2.1. Epilepsy

Figure 2.2: Most frequently reported causes of epilepsy, as reported by countries part of WHO (N=149). Adapted from WHO’s Atlas: Epilepsy Care in the World [18].

the discharges and clinical manifestations may cease in the same abrupt way that they began or there may be a more or less prolonged post-ictal disturbance accompanied by unconsciousness or confusion. In patients with generalized seizures, the discharges are themselves bilateral, symmetrical, and synchronous. They show greater amplitude in the frontal regions but sometimes posteriorly as well [19].

Partial seizures have a localized onset. In this case, consciousness is usually preserved and clinical manifestations are limited to one disturbance, such as, for example, involuntary movements of the arm. The seizure does not have to remain localized, which means involved neurons can recruit neighboring neurons. If, in this process, consciousness becomes reduced, then these seizures are known as complex partial seizures. The activity can spread even more, which will lead to partial seizures with secondary generalization, where consciousness is rarely present [21]. Patients with partial seizures have abnormalities with a topography that corresponds very closely to where the seizure arose [19]. In many patients, the area responsible for seizure generation, or the seizure onset zone (SOZ), can be difficult to specify. Non-agreeing clinical and laboratory studies in patients with identifiable lesions on brain MRI will often indicate poor localization of the SOZ [22].

If an abnormal electrical discharge originates in the motor cortex, the patient will experience a motor seizure; if it takes place in the sensory cortex, it will be a sensory perception; if it happens in the visual cortex, there will be lights, flashes, or jagged lines [20]. If a seizure occurs in the deep temporal lobe there will be a loss of memory or awareness and stop of all activities. The spreading of a seizure to all regions of the brain leads to a tonic-clonic seizure accompanied by loss of consciousness, stiffening and jerking. When persistent seizure activity is present and consciousness (in the case of generalized seizures) is absent for more than 30 minutes, then patients present astatus epilepticus[21].

(32)

2. THEORETICAL FRAMEWORK 2.2. Electroencephalography

Interictal discharges usually occur singly or sporadically. The brain generates its normal rhythms and only now and then interictal epileptiform discharges, limiting the sensitivity of a routine EEG recording. These measurements can be made by surface or intracranial recording; and in short-term (under 30 minutes) or long-term monitoring. The time interval between discharges can vary from minutes to days, which is why in a 20 minute EEG recording, the interictal events may not be present [21, 24]. Intracranial recording is not routinely used due to its invasive nature. A more in-depth analysis of the use of the EEG in the diagnosis of epilepsy can be found in subsection 2.2.3.

2.1.2 Treatment

The goal of treatment is to reduce the likelihood of seizures, ideally to a level where it can be compared to the general population. Options include: treatment with medication -anti-epileptic drugs (AEDs); neurostimulation; and treatment by resection of the epileptic focus [21]. The choice that is made depends on several factors such as the number and severity of seizures that the patient would experience without treatment, the underlying cause, and the age.

Treatment for epilepsy is available and in a majority of cases it can guarantee a normal life. Success of treatment depends on a variety of factors, such as type of seizure, how early the diagnostic is made, the efficacy of medication, compliance with medication, the existence of other associated lesions, and social professional problems. Some epilepsies in children heal always, other types almost always, and only some need permanent anti-epileptic medication. In general, 70 % of patients are free of seizure fifteen years after the beginning of medication [25].

For about 70% of the patients, treatment with medication is suitable; however, for the remaining patients, seizures are not well controlled. In the latter group, if the seizures are focal, with a well defined cortical generator, surgery can often be performed [25].

2.2 Electroencephalography

The outer surface of the cerebral hemispheres, the cerebral cortex, contains neurons (grey matter) and is separated into regions by fissures (sulci). Underneath the cortex are the nerve fibers which connect to other parts of the brain and the body (white matter). Cortical potentials originate from the excitatory and inhibitory post-synaptic potentials generated by cell bodies and dendrites of pyramidal neurons. The scalp EEG represents an average of the electrical activity of a small area in the cortical surface underneath an electrode.

(33)

signal comes form the pyramidal cells. Pyramidal cells make up 70 to 80% of all cortical neurons and their dendrites with the synaptic inputs are perpendicular to the cortex surface. The current dipoles created in individual pyramidal cells are too small to result in a reliably measured electrical signal at the scalp. Synchronous excitatory and inhibitory post-synaptic potentials can arise from the activity synchronization done by the cortical pyramidal cells.

The amplitudes of the scalp EEG range between 10 and 100 µV. The frequency range of the EEG (scalp and intracranial) has diffuse lower and upper limits (about 0.05 to 600 Hz). Due to the fact that there are ultra-fast and ultra-slow frequency components that seem to play no significant role in the clinical EEG, the frequency response curve of an EEG concentrates on the clinically relevant range (0.5 to 35 Hz) [15]. Ultra-slow oscillations may reflect slow cortical potentials that may occur during spreading depression. The predominant frequencies can be divided into the bands seen in table 2.1 with the corresponding wave oscillations in figure 2.3.

Table 2.1:Common definition of frequency bands in the EEG.

Band Frequency Delta Below 4 Hz Theta 4-8 Hz Alpha 8-13 Hz Beta 13-30 Hz Gamma Above 30 Hz

Figure 2.3: Characteristic EEG rhythms - delta, theta, alpha, and beta. As defined in table 2.1. Adapted from Blinowska and Durka [27].

(34)

2.2.1 Historical Background

The history of electroencephalography and epilepsy are closely related. Considering descriptions of seizures in ancient literature such as in Akkadian (oldest written language), ancient Egyptian, Indian, and Chinese; epilepsy can be said to be as old as mankind. The first book on epilepsy was “The Sacred Disease”, where a large portion of text was written by a number of physicians of the Hippocratic School 2400 years ago. It initially was suggested that epilepsy was due to divine influences or magic [28].

Luigi Galvani is the founder of the field of electrophysiology. With the publishing of his discovery of animal electricity, based on the experiment in figure 2.4, he paved the way to understanding epilepsy. His concept of animal electricity - electricity was generated in the body and channeled through nerves - went against the beliefs of contemporary physicist Volta. Volta stated that electricity was generated by plates of different metals. The acceptance of Galvani’s ideas suffered a delay of about three decades due to the dominance of the scientific area by Volta. With the publishing of a book by Du-Bois Reymond, which included an illustrated registration of muscle potential from surface recordings, there was renewed interest in Galvani’s work [28]. With this came the establishment of the basis of clinical electromyography (EMG). As Galvani supposed in the 18th century, animal electricity exists in a state of disequilibrium, and it is, therefore, ready to move in response to any internal stimuli or following external influences [29].

Figure 2.4: Galvani’s experiment that pioneered the subject of electrophysiology. This involved the study of muscular contraction in a frog by touching its nerves with electrostatically charged metal [29].

The electrical activity of the brain, through the intact skull, was actually only first measured in 1923 by Hans Berger. At the time, this measurement was a major accomplishment because there were no modern operational amplifiers available, with their high input impedance. Bioelectric signals were usually measured with a string galvanometer. Nowadays, EEGs are recorded with digital equipment, using high-impedance and low-noise amplifiers. The digital recording offers, among others, the possibility of subsequent signal analysis such as filtering [21].

2.2.2 Technical Aspects

When measuring a biological electrical process that is based on the flow of ions, the ionic currents are converted into electronic currents. At the skin-electrolyte-metal interface, originated by the silver/silver-chloride electrodes and a conducting gel, the conversion of ionic to electronic currents takes place.

(35)

reference electrode or the average signal of all other electrodes [21]. The former is known as common reference and the latter as common average reference.

Figure 2.5: Diagram of recording a single EEG channel. The differential amplifier measures the potential between two electrodes, where one of them is treated as the reference. The second amplifier prepares the signal for AD conversion and storage (lower path). Before the development to digital, EEG was stored on folded paper (upper path). Adapted from Blinowska and Durka [27].

In the process of recording EEG data, by measuring voltage differences between the two inputs, the resulting signal will be amplified (typically 60 to 100 dB voltage gain) and displayed as one channel of EEG activity. The amplified signal will be digitalized via a analog-to-digital (AD) converter, as seen in figure 2.5. Sampling rates range from 256 to 512 Hz in a clinical setting and up to 20 kHz in research applications.

2.2.3 EEG in Epilepsy

The potential for the EEG to identify specific interictal and ictal patterns was first demonstrated by Gibbs et al. [30] in 1935. In the present-day, the scalp electroencephalogram is the most widely accepted test for the diagnosis of epilepsy [12]. Technology has advanced, especially with the introduction of multichannel recordings, prolonged ambulatory records, spectral analysis, video telemetry, and semi-automated analysis of epileptiform activity [31]. However, the EEG has relatively low sensitivity in the interictal period, not showing clear epileptiform abnormalities in 45% of awake EEGs and 20% of sleep EEGs of patients with epilepsy [12, 21]. A vast majority of patients do not have seizures during the somewhat brief EEG recordings, making it difficult to reach a conclusive diagnosis. Also, a physician rarely has the opportunity to observe a patient’s seizure directly, which means that the interictal EEG alterations must suffice in terms of confirmation of the diagnosis of epilepsy but also in classifying the seizure type [24].

(36)

Figure 2.6:24 seconds of EEG around an epilepsy seizure onset for a patient of a study conducted by Lantzet al.[32]. The vertical bar indicates a visually estimated seizure onset.

The specificity and sensitivity of the EEG in patients with epilepsy depend on the type of seizure disorder and the localization of the epileptogenic zone. There can also be attenuation of spike activity by the dura, the bone, and the scalp, therefore altering the diagnostic outcome of the EEG [24]. There are several characteristic EEG patterns associated with well-known and well-defined epilepsy syndromes, as can be seen in table 2.2. EEG can therefore assist in defining certain syndromes, which will influence the decision for therapy and assessment of prognosis [33].

Table 2.2: Typical interictal epileptiform discharges found on the EEGs of patients with characteristic epilepsy syndromes or etiologies [33].

EEG pattern Epileptic syndrome/etiology

Anterior temporal spikes Mesial temporal lobe epilepsy Generalized 3-Hz spike-wave complexes Absence epilepsy

>4-Hz spike-wave complexes, generalized polyspikes Juvenile myoclonic epilepsy Generalized slow spike-wave complexes Lennox-Gastaut syndrome Regional (extratemporal) polyspikes Focal cortical dysplasia

Hypsarrhythmia West syndrome

(37)

seizure discharge or perhaps if the clinical test did not cover the cortical function that is altered during a seizure. The symptomatogenic zone is defined as the area of the cortex which, when activated by epileptiform discharges, produces ictal symptoms. The spreading of the epileptic activity into the symptomatogenic cortex will eventually lead to symptoms [33].

Specific interictal EEG patterns have a high degree of correlation with the presence of epilepsy, which is why about 90% of patients with epilepsy exhibit abnormal discharges during the intervals between seizures [19]. A great majority of patterns include the sharp wave, the spike, and the spike-wave complex [15, 19, 24]. However, patterns may also include benign epileptiform discharges of childhood, slow spike-wave complexes, 3-Hz spike-wave complexes, polyspikes, hypsarrhythmia, seizure pattern, and status pattern [21, 33]. These complexes are much briefer than the ictal discharges [24]. A sharp wave is transient and clearly distinguishable from background activity, having a duration of 70 to 200 ms. A spike is essentially the same as a sharp wave but with a duration of only 50 to 70 ms [21]. The spike and the slow wave are topographically distinct even though the details of their distribution are not exactly clear [19]. In a majority of situations, the slow spike-wave complex consists of a slow spike and a slow wave. Some cases (eg. figure 2.7) consist of true spikes (60 ms or less in duration) followed by a slow wave [15]. The spike-slow-wave complex is a pattern with a spike followed by a slow wave, where the spike has a lower amplitude than the slow wave. It is also possible to have a multiple spike-and-slow-wave complex, which is the same as the spike-slow-wave complex but with two or more spikes associated with one or more slow waves [21].

Figure 2.7:An example of generalized slow spike-wave complexes (around 2 s) found in a child with severe epileptic seizure disorder [15].

It is important to keep in mind the possibility of over interpretation, which can lead to a misdiagnosis of epilepsy [35]. In table 2.3 there is a brief summary of several sharp variants, as well as their characteristics, that can easily be mistaken with epileptiform discharges.

(38)

2. THEORETICAL FRAMEWORK 2.3. Transcranial Magnetic Stimulation

Table 2.3: Sharp transients representing normal EEG variants, being easily confused with epileptiform discharges [33].

Frequency

(Hz) Localization Waveform

Level of

consciousness Age Duration Rhythmical

temporal theta

4-7 Temporal Notched, rhythmic

Relaxed wake sleep stage 1

Young

adults 10 s

Wicket

spike 6-12 Temporal

Monophasic, similar to µ waves

Wake sleep

stage Adults 0.5 s

Small sharp spike sporadic (about 50 ms Frontal maximum Amplitude <50 µV, duration <50 ms Relaxed wake, sleep stages 1 and 2 Adults Single discharges 14- and 6-Hz positive “spikes”

14 and 6

Lateral to posterior temporal

Monophasic Wake sleep stages 1 and 2

Adolescents, adults <1 s

6-Hz “spike and wave” 5-7 Generalized Diphasic, small spike and large wave

Sleep stage 1 Adolescents, adults <1 s

SREDA 5-6 Generalized

Sudden onset and sudden end

Wake sleep stage 1 / hyperventilation

Elderly 40-80 s

At the moment, EEG interpretation in a clinical setting has been based on visual analysis. This analysis usually includes speculative formulation which serves as a guide for investigating the EEG signal and its various graphic elements [36]. A clinical neurophysiologist, trained in the interpretation of the various EEG rhythms, evaluates the waveforms. This evaluation includes assessing the spatial distribution of various frequencies and the reaction to a variety of stimuli, including eyes opening and closing, hyperventilation, and photic stimulation. These aspects contribute to the mean statistical characteristics of the EEG signal, highlighting their importance as background pattern. However, this approach has its drawbacks, such as the long learning curve, the inherent subjective elements, as well as intra- and inter-observer inconsistencies. Due to this, researchers have been motivated into exploring if the computer can assist in extracting relevant EEG features [21, 36]. Yet, it is important to note that the aim is not to replace the classical visual EEG analysis, but quantitative EEG techniques should in fact assist and replace some elements that for now are considered the sole domain of experienced electro-electroencephalographs. The classical visual analysis remains essential for the final interpretation [36].

2.3 Transcranial Magnetic Stimulation

(39)

2.3.1 Historical Background

In 1985, Barker et al. [38] described a novel method for the direct stimulation of the human motor cortex - TMS. Previously, only electrical stimulation had been used, with applications in the human brain and spinal cord [39]. However, due to the activation of the nerve endings in the scalp, this method was quite painful for the subjects and, in many situations, failed to evoke a response. With the existence of the magnetic field induction of currents in TMS, activation of nerve endings does not occur, thus avoiding pain [40]. Nevertheless, electrical brain stimulation is possible today in a non-invasive form and with less discomfort, using scalp electrodes.

The first description of electromagnetic induction was done by Michael Faraday in 1831 at the Royal Institution of Great Britain. His experiment consisted in winding two coils in an iron ring and showing that when the coil on one side was connected or disconnected from a battery, there was an electrical current passing through to the coil on the other side. When the experiment was repeated, a few weeks later, the same effect was produced but this time with two coils closely positioned in air [41]. In fact, if a pulse of current that passes through a coil placed over a person’s head has enough strength and is short enough in duration, there will be a production of rapidly changing magnetic pulses. These pulses penetrate scalp and skull with negligible attenuation, therefore being able to reach the brain [13]. Presently, the stimulating coil acts as the first coil, air is the medium for the magnetic field flow, and the second coil is in fact the electrically conductive living tissue in the area being stimulated [41].

Recordings of experiments related to magnetic stimulation of the brain date back to 1896, when d’Arsonval [42] reported seeing flickering lights in the visual field when he placed his head between two coils with a 110 V supply at 30 A, which involved a direct stimulation of the retina. His report included a description of “phosphenes and vertigo, and in some persons, syncope”. In 1959, magnetic nerve stimulation was accomplished by Kolinet al.[43] in a frog and then in 1965, Bickford and Fremming [44] demonstrated the stimulation of human facial nerves. Due to the long-lasting activation interval, after using an oscillatory magnetic field that lasted 40 ms, it was impossible to record nerve or muscle activation potentials, leading to the non pursuit of the technique for some time. Using 2-ms-duration pulses, Polson, Barker and Freeston [45], in 1982, recorded, for the first time, motor evoked potentials (MEPs). However, in 1985 came the real success as the group made the first clinical examinations with TMS [38].

Since 1985, there have been major improvements regarding equipment reliability and the development of stimulators with differing output waveforms. Coil design, specifically with multiple windings for precise stimulation of nerves or cortical neurons, has been an area of investment [41]. Devices are usually equipped with figure-of-eight coils, which induce a more focused electrical field in the circular coil [1, 13, 46]. This leads to a better control of the excitation produced by the field, and allows a somewhat detailed mapping of cortical representation [13, 46]. The circular coil induces a more widely distributed electric field which allows bi-hemispheric stimulation, important in the study of central motor conduction [13, 47]. An important development in 1988 was repetitive TMS (rTMS), where sequences of stimuli at 1 to 50 Hz [46] are delivered.

(40)

2.3.2 Technical Aspects

TMS is a technique that stimulates the human motor cortex in a pain-free, non-invasive and contactless form [37, 38, 48, 49]. A pulsed magnetic field is applied through the use of a coil which is placed above the subject’s head. A common measurement is placing the coil over the region of interest in the motor cortex and observing movements in the hand or leg on the opposite side of the stimulation [5, 38]. It provides a safe and sensitive measure of both inhibitory and excitatory functions of motor cortical neurons [3].

TMS is defined by the passage of a brief, single, high intensity current pulse in a coil of wire, producing a magnetic field. As the magnetic field penetrates skin and bone it is able to reach the brain with negligible attenuation, and creates an electric field [13, 14, 37, 48]. The physical foundation of TMS can be described by Maxwell’s equations. The time-varying current pulse in the stimulation coil will produce a magnetic field, in accordance with the Biot-Savart law. The time-varying magnetic field will induce an electric field, following Faraday’s law. This induced electric field will move charges in the direction of its field lines. The coil can be parallel to the surface of the conductor (in this case, the head) or not. Depending on its position, surface charges will appear due to induction or they will accumulate at the conductor surface and in interfaces between tissues with different conductivity, generating a secondary electric field [50].

The total induced electric field inside a conductor (E) is better represented by the general written form of Faraday’s law, where the left side of the equation mathematically describes the curl (∇~_{) of the electric field (}_E_{) and the right side represents the rate of}

change of the magnetic field (B) over time.

~

∇ ×_E₌−∂B

∂t (2.1)

The pulse-generating circuit of the magnetic stimulator produces monophasic or damped sinusoidal (biphasic) current pulses. The decaying current oscillation (I), because of resistive losses in the circuit, obeys the form in equation 2.2.

I(t) =

U0

Lω

e−(

R

2L)tsin(ωt) (2.2)

whereω=

q 1 LC − R 2L 2

,Cis the capacitance,U₀is the capacitor’s initial voltage,Lis the inductance of the coil, andR is the common resistance of the components in the circuit represented in figure 2.8.

(41)

No contact to the head is necessary; the scalp and the skull have almost no effect on the magnetic field [1]. However, due to the fact that the strength of the magnetic field falls off very rapidly with distance (square of the distance from the stimulating coil) [41], it will only penetrate a few centimeters, meaning that only superficial areas of the brain are most effectively stimulated [14, 50]. The electrical currents will depolarize cell membranes so that voltage-sensitive ion channels are opened and action potentials are initiated [51]. The cortex is activated to a depth of two to three centimeters and with a surface area of several square centimeters, considering the commonly used stimulation intensity and coils (see figure 2.9) [52]. Excited neural structures [48] will then stimulate muscles, peripheral nerves and cortical neurons [14] without requiring surgical access or anesthetic agents [1]. Its uniqueness relies on the fact that it activates all its primary target neurons at the same time [51].

Figure 2.9: Principles and chain of events in TMS. The current pulses in the coil generates a magnetic fieldBthat, in turn, induces an electric fieldEthat is strongest near the coil. The electric field aligns tangentially to the head surface (closed circles) The pyramidal axons depolarize at their bends, affecting transmembrane potential, and consequently leading to the firing of the neuron. Scalp recorded EEG reflects synchronous activity evoked by TMS [46, 50].

Cortical stimulation can activate, inhibit or interfere in other ways with the activity of cortico-subcortical networks [48]. This will depend on the stimulus frequency and intensity, current polarity, coil orientation, and the configuration of the induced electric field [48, 51].

(42)

The coil can be moved until the adequate stimulation site is reached [38]. This is important because the currents generated by TMS and their physiological effects can be modulated by coil construction and positioning, and even by brain conductivity and neuronal orientation. Distribution of field strength and flux is difficult to predict [49]. The overall response amplitudes are highest right underneath the coil, decreasing as the distance from the stimulation point increases [51]. Also, stimulation of the motor cortex with different current directions in the circular coil will yield different responses [53].

There are several parameters that can be measured with the use of TMS, these include motor threshold (MT), motor evoked potential (MEP) amplitude, stimulus-response curve, “phosphene threshold”, cortical silent period, intra-cortical inhibition, and intra-cortical facilitation. These are addressed in table 2.4.

Table 2.4: TMS parameters that can be obtained by its use in several forms depending on the objective of the experiment [13, 49, 54]. The physiological significance of measuring these parameters is also shown.

Parameter Measurement Physiological significance

Motor threshold Single pulse: it is the threshold for motor response

Cortical neuronal membrane excitability; corticospinal system threshold excitability

MEP amplitude Single pulse: average of maximal amplitude

Excitable proportion of neuronal pool

Stimulus-response curve

Single pulse: refers to the increase in peak-to-peak MEP amplitude as a function of TMS intensity

Assesses the neurons that are away from the core region which is activated at MT

“Phosphene threshold” Single pulse: it is the threshold for visual response

To study the occipital cortex and the visual pathways

Cortical silent period

Single pulse: observation of reduced post-MEP background activity during muscle contraction

Cortical inhibitory mechanisms

Intracortical inhibition

Paired subthreshold conditioning and suprathreshold pulses 2- to 5-ms delay

Possibly GABAergic

Intracortical facilitation

Paired subthreshold conditioning and suprathreshold pulses 7- to 20-ms delay

Uncertain

2.3.2.1 Motor Evoked Potentials

Motor threshold refers to the lowest TMS intensity necessary to evoke MEPs in the target muscle when single pulse stimuli are applied to the motor cortex as mentioned in table 2.4. The MT should relate to the activity of neural inputs into pyramidal cells that will ultimately affect their membrane excitability. This provides an insight into the efficacy of a chain of synapses from pre-synaptic cortical neurons to muscles. There is an alteration in the threshold if a certain disease affecting the pathways from neurons to muscles, is present [13]. EEG activity at low TMS intensities, which means below MT, probably has different distributions than at higher intensities [55].

(43)

Selecting targets for TMS according to anatomical brain structures in different subjects does not always lead to the stimulation of the same area. This is due to the inter-individual differences in structure-function relationships, in other words, between brain anatomy and functional architecture [51]. The variable amplitude of the muscle response to TMS of the motor cortex is due to the easiness in producing large MEPs in some healthy subjects. In others, the cortico-muscular pathways can be barely excitable, therefore producing low MEPs. It is, therefore, important to keep in mind that differences found among healthy subjects and patients are important sources of data.

Disparities between individuals include age, genetic factors, physiological differences associated with behavior, and other traits. Intra-individual variances are strongly influenced by time and external factors [59]. Experimental groups should attempt a construction as balanced as possible in sex and demographic factors, such as age and education. Thus, the different coil positioning approaches do not necessarily imply a huge qualitative difference in the TMS-induced effect, but in the magnitude of the respective effect size [60]. Responses therefore depend on the exact coil location and orientation, on the state of the cortex and on the state of vigilance of the subject [8, 61,62].

2.3.3 TMS in Epilepsy

TMS delivered at different levels of the motor system can provide information regarding cortical excitability; the functional integrity and efficacy of area-to-area neuronal connections [1]; the conduction along corticospinal, corticonuclear, and callosal fibres; the function of nerve roots and peripheral motor pathway to the muscles [13]. It is also possible to perturb on-going neuronal signal processing in the brain with the purpose of finding cortical areas that are important for specific tasks. With this it becomes possible to treat patients using repetitive stimulation by targeting specific cortical areas [1]. This can help localize the level of a particular lesion within the nervous system or even to predict the functional motor outcome after an injury. An important aspect is the fact that the abnormalities revealed by TMS are not disease-specific, so results should be interpreted with other clinical data. Some findings can be useful for an early diagnosis and prognostic prediction [13].

Since TMS is a measure for excitability (i.e. how easily a response can be evoked) it becomes clear how it should be applied in the diagnostic process of epilepsy. As previously described, epilepsy is characterized by an increase in excitability. With a tool to measure this increase, there has been a rise in the studies that are investigating the use of TMS in epilepsy research.

Applications include investigation of the underlying cortical excitability, determination of the effects of AEDs, pre-operative localization of the epileptic foci and even functional mapping [49]. The ability of a short-lasting magnetic field inducing an electrical current within body tissue allows the researcher to influence or monitor the neuromuscular system. It can also be used to influence sensory neurons in the brain [41]. TMS is an attractive tool for the study of seizure disorders due to its simplicity; it is relatively inexpensive and generally safe. So far, results obtained from TMS studies suggest that patients with generalized epilepsy syndromes have increased cortical excitability, which makes this technique an adequate mean for clinical and research applications [49]. As a diagnostic tool, single-pulse and paired-pulse TMS may be used to map cortical function and also to measure cortical excitability [63].

(44)

2. THEORETICAL FRAMEWORK 2.4. TMS-EEG

patients not undergoing any treatment [54]. A recent study by Badawyet al.[3] indicates that AEDs suppress seizures by modulating their cellular target in a ways as to change the pattern of the pre-existing cortical hyperexcitability in epilepsy. This study involved the use of TMS, like previous studies before it, with the objective of studying the effect of prolonged AED use and how it affects cortical excitability in epilepsy.

2.4 TMS-EEG

Until a few years ago, most TMS experiments and applications were limited to the stimulation of the motor cortex, because the only observable effects were those that reflected peripheral muscular activity [1]. However, in the past years, it has been demonstrated that the effects of TMS can also be observed by means of SPECT, NIRS [2], functional MRI, PET, and EEG [1] in a more direct manner. The first four techniques have poor temporal resolution because they use the variation in blood flow and oxygenation to detect changes in neuronal activity. Since EEG directly measures the electrical activity of neurons, it has an excellent temporal resolution.

The ability of the EEG to measure direct cortical activation that is induced by TMS shows the importance of using EEG and TMS simultaneously. Unlike any other available brain imaging method, the EEG is able to provide a mean to study the instantaneous neuronal effects of TMS in the brain, and thus probe the brain’s excitability [2, 51]. This can provide information regarding the state of the stimulated area as well as the functional connectivity to other regions and their state [55]. TMS-EEG can access any cortical region (primary and associative) in any category of patients, providing a straightforward and flexible way to monitor the state of corticothalamic circuits [64]. Detection and monitoring of the state of corticothalamic circuits therefore becomes more straightforward and flexible [51]. A variety of information can be obtained by altering the TMS intensities, inter-stimulus intervals, induced current direction, and cortical targets [55].

The temporal resolution of TMS is, in theory, only limited by the duration of the TMS pulse (about one millisecond). In the EEG the temporal resolution is limited by the sampling frequency. However, the combination of the two techniques is not determined only by their nature, but also by their interaction. It can take the amplifier several milliseconds to reset after the TMS is applied and the emergence of neural activity, that can generate a detectable signal, can also take some time. EEG is sensitive not only to the firing rates of the underlying neural activity but also to the synchrony of the activity, and the geometry of the active neural elements [37], which means there can be a slight delay in obtaining a signal.

2.4.1 Analysis Methods

(45)

the TMS pulse also activates the muscles in the underlying region of the scalp for a short period of time, creating a light twitching sensation. The rapid movement of the component wire within the coil will cause a loud click, heard every time a pulse is given [37]. A way to deal with the artifact problem is to exclude the channels that are strongly affected by said artifact. However, there is a problem with this solution because these channels are usually the ones closest to the stimulation site, thus they are usually the most informative about the early stages of response [5].

In the use of multi-channel EEG recordings, it is necessary to start artifact removal and data analysis during acquisition. This requires appropriate technological solutions for the recording environment, electrodes, amplifiers, a careful methodological approach, and suitable analysis methods should be used to eliminate de effects of the TMS pulse. The pulse is strong enough to cause significant and visible disturbances in the EEG [51]. However, filters should not be used during recording because these interact with the residual spike-shaped artifact which leads to a ripple in the signal after each TMS pulse that can last up to one second. Filters can be used after recording, once the discharge artifacts from the TMS have been removed [37].

There are several reasons that can account for the long-lasting TMS artifact. The major influence could be due to the fact that electrodes and skin have magnetic properties, which may therefore be affected by the TMS pulse and generating an extra-cortical signal in the recording [65]. Re-positioning of the coil or even due to head movements in one experiment can be a source of artifact. Reasons could include the fact that while the coil position is optimized by examining motor evoked responses, the angle of the coil with respect to the electrodes depends on head size and shape, and local skull curvature under the coil. Even the smallest difference in coil orientation can have major effects on the effective magnetic field strength near a specific electrode [5]. On the other hand, perhaps the exact angle of the electrode with respect to the spatial gradient of the field can also make a difference in the amount of charge that can be accumulated at the skin-electrode junction as a result of the TMS pulse [5].

2.4.2 Responses

EEG can be used to locate the neuronal activity evoked by TMS, and how it spreads to other regions, in order to determine reactivity and connectivity patterns [2]. It is also possible to develop studies about how the brain processes information from the periphery, by determining, temporo-spatially, the effects of TMS on evoked potentials (EPs) and event-related potentials (ERPs). EEG can also be used to monitor abnormalities or to control the efficacy of the use of TMS as a treatment [46].

ERPs are the measured brain response by EEG to external stimuli. Short-latency ERPs are influenced mostly in the physical characteristics of the stimulus while longer-latency ERPs predominantly depend on the conditions of how the stimuli is presented [34]. The causality between the detected activations is evident through their temporal sequence, as long as the EEG’s temporal resolution is sufficient to identify the neural phenomena [2]. An important aspect of TMS-evoked EEG topography is that even though only one cortical hemisphere is stimulated, bilateral EEG responses can be evoked and with different features because of inter-hemisphere connections [51].

(46)

A typical scalp-recorded averaged TMS-evoked EEG signal can be seen in figure 2.10. There are several deflections, first as rapid oscillations and then as lower frequency waves. The responses depend on the state of the cortex in that instant [61, 62] and on the location of the stimulation [8]. The TMS-evoked average responses are usually highly reproducible, as long as the delivery and targeting of TMS is controlled and stable from pulse to pulse and between experiments [51].

Several peaks are identifiable in the typical TMS-EEG response in figure 2.10, such as P30, N45, P55, N100, and P180. These represent the time, in milliseconds, after the TMS pulse, at which they occur. While N stands for a negative peak, P is a positive peak. The N100 is the most pronounced, reproducible and long-lasting component in response to motor cortical TMS, according to reports [7, 62, 66]. There can be some small deviations in the time at which these peaks occur. Recording EEG during TMS can be a technically challenging task because TMS induces a very strong electrical field which could saturate recording amplifiers for quite some time [65].

Figure 2.10: TEP: TMS-evoked potential. There is a clear identification of the major peaks and their polarities, P30, N45, P55, N100, and P180. P=positive, N=negative and the number represents the time at which said peak occurs in milliseconds. This is a single-channel response. The structure and latency of these peaks may vary between subjects and measurements [51]

The sub-millisecond synchronization, observed initially, is soon lost due to the conduction from the site of stimulation to the first synapses and further along the neuronal network, initiating a cascade of serial and parallel effects. The stimulated cortex assumes an inhibitory state for a period of 100 ms or more, because of the activation of inhibitory cells as well as excitatory cells [51]. This is known as the cortical silent period, evidenced by a period of EMG silence following each MEP when the subject tries to maintain spontaneous muscle activity during the whole measurement. Most of the silent period is believed to be due to inhibitory mechanisms at the motor cortex [13, 67].

Even though there have been many studies regarding simultaneous TMS and scalp EEG [6–9], none addressed a comparison between healthy controls and individuals suffering from a neurological condition. In the study by Valentínet al.[12], EEG responses to TMS are described as well as how they can be used to evaluate focal epilepsy. This evaluation can be for diagnostic purposes or to identify the epileptogenic cortex during presurgical assessment. Prior to this study, TMS-EEG responses had not been evaluated for the diagnosis of epilepsy.

(47)

populations, or even as component of a responsive neurostimulation set up in which TMS timing is determined by underlying EEG activity [68]. It could also become an alternative method to identify epileptogenic cortex non-invasively in patients with epilepsy [12].

Valentínet al.[12] saw several types of responses to TMS in the EEG. Early responses are in seen in both groups of individuals being studied: patients and healthy controls. With this outcome, the focus was given to late responses, where they saw a difference. Their results suggested that the use of TMS can increase the diagnostic sensitivity of the EEG in epilepsy. What they defined as delayed responses appear to be equivalent to the ones that were described in their previous studies by patients with intracranial electrodes. Due to the fact that late TMS-EEG responses were seen in zero of the 15 healthy subjects and in 11 of the 15 patients, they consider these responses as abnormal. This could be related to the hyperexcitable cortex existing between the stimulated area.

(48)

Quantification of the TMS-EEG response in epilepsy

Maria Inês Fonseca Silva Santos